Computer performance is a key success factor in virtually all fields of research that Siemens is involved in. Today’s computers, which work with binary codes, are not ideally suited to many calculation tasks, which is why a quantum computer would offer a much better option for the future. Experts believe that such a computer would be much faster than today’s units in terms of its ability to recognize patterns in many applications such as image processing, detecting viruses, and the analysis of genetic databases. A quantum computer would also be able to reliably read hand-written addresses on envelopes and more effectively monitor technical facilities.

In cooperation with Munich Technical University, researchers from Corporate Technology have now taken a giant step toward improved information processing with quantum computers by successfully completing the first-ever experiment to create an artificial neural network on a simple quantum computer.

Specialists from CT’s Learning-Enabled Systems department have been working with artificial neural networks for many years now. Such networks operate in a manner similar to that of the human brain and are especially suited to pattern recognition operations. They are able to learn and can be trained via examples. The idea behind placing neural networks on quantum computers is to ensure more efficient processing of the huge amounts of data associated with pattern recognition. Instead of bits, quantum computers work with data units known as quantum bits, or qubits. These units are capable of assuming different states simultaneously, and can also be entangled with other qubits in a special type of quantum correlation. Because of these properties, computer calculations with qubits are much faster — and more complex — than operations with conventional bits.

In his Siemens-sponsored doctoral dissertation, quantum computer programmer Rodion Neigovzen simulated a complete system consisting of a quantum computer and a neural network. He then created a program to run on it that can compare a bit pattern consisting of various colors with stored sample patterns, and subsequently calculate the degree of similarity between them. Researchers at Munich Technical University then worked closely with Neigovzen to carry out a feasibility study for the system in an NMR spectrometer. Here, a room-temperature solution of sodium formate was used. Among other things, this compound contains one carbon and one hydrogen atom. In strong magnetic fields, the nuclear spins of both particles each form one qubit with two possible states. The quantum computer signals measured in the feasibility study corresponded extremely closely to the signals calculated and postulated by Neigovzen, thereby confirming that the researchers’ algorithm for a quantum computer delivers accurate results in practice.